| چکیده انگلیسی مقاله |
Suspended sediment concentration (SSC) is one of the most important factors in association with estimation and monitoring of flood characteristics. SSC represent water pollution and sediment characteristics in upstream of a river and reservoirs and dams characteristics in downstream of a river. In this study, deposited suspended sediments in the bottom of the stream were collected and dried. The dried suspended sediments were then dissolved in the measured volume of water. Then, 332 images were taken based on suspended sediment different concentrations in the range of 0.0005 to 200 mgr.l-1 using a mobile camera. Finally, the average Red, Green and Blue (RGB) bands values and also, Hue, Saturation, and Brightness (HSV) were extracted for each image based on image processing technique using the Matlab 2020 software and Python programming language in Google Colab environment. In this matter, Artificial Neural Networks (ANN) models based on training algorithms viz., Bayesian Regularization backpropagation (BR), Conjugate Gradient backpropagation (CG) and Levenberg-Marquardt backpropagation (LM) were applied to evaluate SSC observation values as dependent variables and images Red, Green and Blue (RGB color model) and Hue, Saturation and Value (HSV color model) as independent variables. The results showed that the coefficient of determination (R2) value for training phase was obtained 0.67 based on Levenberg-Marquardt algorithm (LM) related to RGB. However, these were 0.56 for HSV. It was concluded that of image processing technique and ANN models could be acceptable method to indirect estimate SSC in the streams in association with reduction of costs and time. |
